The following document provides an initial descriptive analysis and data coverage overview of the [Annual Survey of State and Local Government Finances. The data is provided in a standardised format from Willamette University. The data is available at a county level between 1970 - 2020 with improved coverage in more recent years. The data has been cleaned and grouped into 17 expenditure categories (listed in data tree below), 4 higher-level categories (comprised of the 17 lower-level categories), and as a county-level total expenditure. All plots are presented using the raw data, whereas the tables presenting the linear time trend results include results for both the raw data and linearly interpolated data.
The data is currently structured in county-year pairs with observations for 17 public expenditure categories. I have manually grouped these 17 categories into 4 higher categories (that can be changed/redefined easily!). The tree structure below displays all categories and groupings. As currently thought out, any or a combination of these categories will serve as outcome/dependent variables.
## levelName
## 1 Total Public Expenditure
## 2 ¦--Commercial
## 3 ¦ ¦--Air Transportation
## 4 ¦ ¦--Liquor Stores
## 5 ¦ °--Port Facilities
## 6 ¦--Public Welfare, Services & Facilities
## 7 ¦ ¦--Education
## 8 ¦ ¦--Fire Protection
## 9 ¦ ¦--Health & Hospitals
## 10 ¦ ¦--Housing & Community Development
## 11 ¦ ¦--Public Services & Facilities
## 12 ¦ ¦--Public Transit
## 13 ¦ ¦--Public Welfare
## 14 ¦ °--Retirement
## 15 ¦--Administration & Misc.
## 16 ¦ ¦--General Administration
## 17 ¦ ¦--Interest on Debt
## 18 ¦ °--NEC
## 19 °--Infrastructure, Utilities & Regulation
## 20 ¦--Highways
## 21 ¦--Police, Corrections & Judicial
## 22 ¦--Protective Inspection & Regulation
## 23 ¦--Utilities
## 24 °--Waste Management & Sewerage
The below plots provide aggregated geographical (national, state, regional) summaries of expenditure categories and shares to demonstrate the composition of national expenditure across the categories. The final two plots demonstrate the composition of expenditure within the Public Welfare, Services & Facilities category.
One of the issues reported by the data collectors is that the surveys are only mandatory every 5 years. So, although data exists and is reported annually, there are several counties that do not report figures in many years. 30% of the panel values are missing due to this. I am unsure what to do with this information and whether there might be any fixes. Climate Econometrics uses indicator saturation to fill in missing values which I would be interested to try but I do not yet know if this is appropriate for this dataset (especially on outcome variables). In the below charts, I have provided a map that shows the proportion of each county’s time series (1970-2020) that is missing. States outlined in red are top fossil-fuel producing states (Texas, Pennsylvania, Oklahoma, Wyoming, West Virginia, North Dakota, Nevada, New Mexico, Louisiana, Colorado).
The following tables provide standard linear time trends evaluated in
a county-fixed effects panel linear regression model for a panel of
3,058 counties and for a time period of 1970-2020.
Dependent variables: Expenditure by category per capita
values.
Data: Raw data and linearly interpolated data (likely a flawed
method).
Result: Per capita expenditure does follow an increasing trend albeit very small and not in line with growth in personal income rates (Pinc K below).
Note: Amount pc denotes total public expenditure and Pinc K denotes personal income per capita (as GDP indicators are not available for the full duration of the panel).
| Outcome Variable |
|
|
|---|---|---|
| Commercial | 0.07*** | 0.08*** |
| 0.01 | 0.01 | |
| Public Welfare Services Facilities | 1.22*** | 1.78*** |
| 0.04 | 0.05 | |
| Administration Misc | 1.42*** | 2.18*** |
| 0.06 | 0.07 | |
| Infrastructure Utilities Regulation | 1.44*** | 2.25*** |
| 0.04 | 0.04 | |
| Pinc K | 878.69*** | 880.19*** |
| 4.37 | 4.39 |
| Outcome Variable |
|
|
|---|---|---|
| Amount Pc | 33*** | 31.72*** |
| 0.63 | 0.6 |
The below shows spatial distribution of trends in personal income,
total direct expenditure, and expenditure (all in per capita values) in
our four broad categories (trend values calculated in a linear time
series model). These graphs show variation in the linear trends across
counties.